Fitting Gammachirp filter model parameters to perceptual second formant data of voiced speech

We propose in this paper a new estimation technique of the perceptual second formant, based on an auditory-filterbank. The auditory fllterbank is simulated by the Gammachirp filter model centred on different frequencies; as the repartition of the hair cells on the basilar membrane, which are responsible of the frequency sensation. The proposed technique achieves a spectral estimation at the output of the filterbank. To validate this technique, an estimation of the second formant of the 18 synthetic cardinal vowels, analyzed by Bladon and Fant, is applied. The obtained results give minimum errors and accurate estimation values of perceptual second formant as Bladon and Fant calculation model and more accurate than other classical methods like those proposed by Carlson and Hermansky models.